Experimental and Theoretical Investigation of the Crossover from the Ultracold to the Quasiclassical Regime of Photodissociation
I. Majewska, ^{1} S. S. Kondov, ^{2} C.H. Lee, ^{2} M. McDonald, ^{2}^{,} ^{∗} B. H. McGuyer, ^{2}^{,} ^{†} R. Moszynski, ^{1} and T. Zelevinsky ^{2}^{,} ^{‡}
^{1} Quantum Chemistry Laboratory, Department of Chemistry, University of Warsaw, Pasteura 1, 02093 Warsaw, Poland ^{2} Department of Physics, Columbia University, 538 West 120th Street, New York, NY 100275255, USA
At ultralow energies, atoms and molecules undergo collisions and reactions that are best described in terms of quantum mechanical wave functions. In contrast, at higher energies these processes can be understood quasiclassically. Here, we investigate the crossover from the quantum mechanical to the quasiclassical regime both experimentally and theoretically for photodissociation of ultracold diatomic strontium molecules. This basic reaction is carried out with a full control of quantum states for the molecules and their photofragments. The photofragment angular distributions are imaged, and calculated using a quantum mechanical model as well as the WKB and a semiclassical approximation that are explicitly compared across a range of photofragment energies. The reaction process is shown to converge to its highenergy (axial recoil) limit when the energy exceeds the height of any reaction barriers. This phenomenon is quantitatively investigated for twochannel photodissociation using intuitive parameters for the channel amplitude and phase. While the axial recoil limit is generally found to be well described by a commonly used quasiclassical model, we ﬁnd that when the photofragments are identical particles, their bosonic or fermionic quantum statistics can cause this model to fail, requiring a quantum mechanical treatment even at high energies.
I.
INTRODUCTION
Lowtemperature atomic and molecular collisions and reactions are qualitatively diﬀerent from their high energy counterparts. While the latter can usually be explained quasiclassically, the former are strongly in ﬂuenced by the wave nature of the particles and are correctly described only with quantum mechanical wave functions. In addition, at very low temperatures, precise control of internal molecular states is possible, permit ting studies of statespeciﬁc reaction cross sections rather than statistical averages. Ultracold, quantum mechanical chemistry has many distinguishing features such as reac tion barrier tunneling, quantum interference, and possi bilities for controlling the outcome with applied electric or magnetic ﬁelds [1]. On the other hand, a quasiclassical interpretation can oﬀer intuitive insight into complicated processes. In particular, photodissociation [2–4] has been exten sively used to study the nature of molecular bonding, which becomes encoded in angular distributions of the outgoing photofragments. It is one of the most basic chemical processes, and is highly amenable to quantum state control of the outgoing particles, since the reaction proceeds without a collision. The majority of photodis sociation experiments to date have been carried out in a regime that is well described in the quasiclassical frame work [5–8], while recently we studied this reaction in the purely quantum regime to observe matterwave interfer
^{∗} Present address: Department of Physics, University of Chicago, 929 East 57th Street GCIS ESB11, Chicago, IL 60637, USA ^{†} Present address: Facebook, Inc., 1 Hacker Way, Menlo Park, CA 94025, USA ^{‡} tanya.zelevinsky@columbia.edu
ence of the photodissociation products [9] and control of the reaction by weak magnetic ﬁelds [10]. In this work, we show that the photodissociation energy can be tuned by several orders of magnitude to span the range from the quantum mechanical to the quasiclassical regime. Ex perimentally and theoretically we demonstrate how the crossover occurs and what determines its energy scale. We also ﬁnd that if the photofragments are identical bosonic or fermionic particles, then the reaction outcome is not always well described in quasiclassical terms even at high energies and the eﬀects of quantum statistics can persist indeﬁnitely. Moreover, we apply the Wentzel KramersBrillouin (WKB) approximation to photodisso ciation, as well as a related semiclassical approximation, and discuss how accurately they capture the transition from ultracold to quasiclassical behavior.
In the experiment, we trap diatomic strontium molecules ( ^{8}^{8} Sr _{2} ) at a temperature of a few microkelvin, enabling precise manipulation and preparation of speciﬁc quantum states as the starting point for photodissocia tion. The photodissociation laser light has strictly con trolled frequency and polarization, yielding photofrag ments in fully deﬁned quantum states. The ab initio theory is aided by a stateoftheart molecular model [11, 12] that captures the eﬀects of nonadiabatic mixing and yields excellent agreement with spectroscopic mea surements [13–16].
This approach oﬀers a unique opportunity to ﬁnely scan a large range of energies that is relevant to the crossover from distinctly quantum mechanical to quasi classical photodissociation. Since initial molecular states have low angular momenta, the relevant dynamics occurs at very low, millikelvin energies. The experiment can sample energies from ∼ 0.1 mK limited by the molecule trap depth to ∼ 100 mK limited by the available laser in tensity that is needed to overcome the diminishing tran
2
sition strengths. The rotational and electronic potential barriers that are explored in this and related [17] work fall within this range, permitting direct observations of the crossover. We reduce the process to a single initial molec ular state and two angularmomentum product channels, in which case the quantum mechanics of the reaction can be encoded in a single pair of coeﬃcients: the amplitude and phase of the channel interference. Note that in both quantum and quasiclassical regimes we work with single initial quantum states. The classical nature of the process emerges not because of statistical averages over the initial states, but when the kinetic en ergies of the photofragments are large. At high energies, the photodissociation time scale is much faster than the time scale of molecular rotations, and the molecules can be described as classical rotors. An equivalent view is that the photodissociation outcome should have quasi classical behavior if the photofragment energy in the con tinuum exceeds the height of any potential barriers. This regime is referred to as axial recoil, where the molecule has insuﬃcient time to rotate during photodissociation and the photofragments emerge along the instantaneous direction of the bond axis. Here we investigate at what energies the axial recoil regime is reached, whether it is accurately described by a quasiclassical model, and how its onset depends on the molecular binding energy and angular momentum. In Sec. II of this work, we describe the photodissoci ation experiment and show the measured and calculated photofragment angular distributions. In Sec. III, we discuss the quantum mechanical photodissociation cross section used in calculating the angular distributions, its axial recoil limit, and the quasiclassical model. Section IV details the crossover from the ultracold to the quasi classical regime for a range of molecular binding energies and angular momenta and shows the onset of the axial recoil limit at energies that exceed the potential barriers in the continuum. In Sec. V we discuss the breakdown of the quasiclassical picture of photodissociation in case of identical photofragments and present the correct high energy limit that takes quantum statistics into account. We summarize our conclusions in Sec. VI.
II.
THE PHOTODISSOCIATION EXPERIMENT
The starting point for the experiment is creation of ∼ 7, 000 weakly bound ultracold ^{8}^{8} Sr _{2} molecules in the electronic ground state. This is accomplished by laser cooling, trapping, and photoassociating Sr atoms in a faroﬀresonant optical lattice [18]. Subsequently, optical selection rules allow speciﬁc molecular quantum states to be populated and serve as a starting point for photodis sociation. Figure 1 shows the photodissociation experiment from the point of view of molecular potentials and wave func
tions. The ground state gerade potential X0
+
g
correlates
to the ^{1} S + ^{1} S atomic threshold, while the singlyexcited
FIG. 1. Molecular potentials for the ^{8}^{8} Sr _{2} electronic states
used in this work. The ground state X0
g
+
correlates to the 

+ 
and 
u 
^{1} S + ^{1} S atomic threshold while the excited states 0
1 _{u} correlate to ^{1} S + ^{3} P _{1} . The 1 _{u} potential has a ∼ 1 mK
electronic barrier. The molecules are photodissociated via an
electricdipole optical transition from bound states of mostly
0
+
u
or 1 _{u} character to the groundstate continuum. A range
of rovibrational initial states are explored. In this example,
weakly bound 0
+
u
(v = −2, J _{i} = 1, M _{i} = 0) molecules are
photodissociated, and the photofragments occupy two allowed partial waves J = {0, 2} shown at the energy of 50 MHz (2.4 mK).
ungerade potentials 0
+
u
and
1 _{u}
correlate to
^{1} S + ^{3} P _{1} ,
where the 0 and 1 state labels correspond to Ω, the pro
jection of the total atomic angular momentum onto the molecular axis. Weakly bound groundstate molecules
are excited to 0
+
u
and 1 _{u} bound states with a laser pulse,
and the resulting molecules are immediately photodisso ciated to the groundstate continuum. The bound states are labeled by their vibrational quantum number v (neg ative numbers count down from the threshold), total an gular momentum J, and its projection M onto the quan
tization axis that is set by a small vertical magnetic ﬁeld.
In Fig. 1, the 0
+
u
(v = −2, J _{i} = 1, M _{i} = 0) molecules are
photodissociated by 689 nm light. The allowed angular momentum states, or partial waves, in the continuum are J = {0, 2} due to electricdipole selection rules as well as the bosonic nature of ^{8}^{8} Sr that allows only even J values in the ground state.
The photodissociation laser light copropagates with the lattice laser and both are focused to a ∼ 30 µm waist at the molecular cloud. A broad imaging beam, resonant with a strong 461 nm transition in atomic Sr, nearly co propagates with the lattice and is directed at a charge coupled device camera that collects absorption images of the photofragments. A typical experiment uses ∼ 10 µs long photodissociation laser pulses, ∼ 100 µs of free expansion, and ∼ 10 µs absorption imaging pulses [17]. Several hundred images are averaged in each experiment.
The experimental and theoretical results of photodis sociating the molecules in a range of precisely prepared quantum states are shown in Fig. 2. In this set of measurements, light polarization is along the quan
3
FIG. 2. Measured and calculated photofragment angular distributions in the ultracold quantum mechanical regime and in the
highenergy axial recoil limit. (a) The 0
+
u
initial states are explored at the continuum energies ε/h = {13, 33, 53, 100, 300}
MHz (ε/k _{B} =0.614 mK). For each initial state, the top and bottom rows correspond to measurements and quantum mechan ical theory, respectively. The ab initio images were adjusted for the shell size and thickness in experimental data, where the photofragment ring diameter is typically 0.30.5 mm. The strong dependence of the patterns on ε and the less pronounced dependence on v are discussed in the text. The lowest row shows vindependent distributions obtained with the WKB approx imation. The most weakly bound state exhibits central dots that arise from spontaneous photodissociation into the optical
lattice; these are more apparent at higher energies where the signal is weaker, and can be ignored. (b) Same as (a), for 1 _{u}
initial molecular states. (c) Angular distributions calculated for a pair of 0
+
u
and 1 _{u} weakly bound states using both quantum
theory and the WKB approximation at a high energy ε/h = 1000 MHz (ε/k _{B} = 48 mK), to show their close agreement with
the appropriate axial recoil limit.
tization axis and the molecules start from J _{i} = 1,
M _{i}
=
0.
In Fig.
2(a), molecules in vibrational states
v = {−2, −3, −4, −5} of 0
+
u
are photodissociated. The
photofragment angular distributions show a strong en
ergy dependence, and a less pronounced dependence on
v except for the most weakly bound levels, as discussed
in Sec. IV A. The bottom row of theoretical images is
obtained with the WKB approximation, which is inde
pendent of v and described in Sec. IV B. Figure 2(b) is analogous to (a) but illustrates 1 _{u} initial states. A good agreement between measurements and quantum mechan ical calculations is reached for all initial states and con tinuum energies. Figure 2(c) shows agreement between quantum theory and the WKB approximation at the con tinuum energy ε/h = 1000 MHz (ε/k _{B} = 48 mK) where h = 2π and k _{B} are the Planck and Boltzmann con
4
stants. At this high energy, both methods approach the
axial recoil limit which is also shown.
A. General quantum mechanical model
The general quantum mechanical expression for the
photodissociation cross section is based on Fermi’s golden
rule with the electricdipole (E1) transition operator,
σ _{Q}_{M} (θ, φ) ∝
JM
k
({r}, R) ^{} T
E1
ˆ
1
^{} Ψ _{J} _{i} _{M} _{i} _{Ω} _{i} ({r}, R) ^{} ,
2
(1)
III.
PHOTOFRAGMENT ANGULAR DISTRIBUTIONS
Sections III AIII E overview the quantum mechanical
calculation and parametrization of the photodissociation
cross section both in the lowenergy regime and in the
highenergy axial recoil limit, as well as the correspon
dence of the latter to the quasiclassical approximation.
where {r} is the set of atomic electronic coordinates, R
is the vector connecting the photofragments, k is the lab
frame wave vector of the photofragments, Ψ
JM
k
is the ﬁ
nal (continuum) wave function, and Ψ _{J} _{i} _{M} _{i} _{Ω} _{i} is the initial
(bound) wave function. The polar and azimuthal angles
{θ, φ} are referenced to the quantization axis.
Assuming a pure initial state J _{i} M _{i} Ω _{i} , expanding the
bound and continuum wavefunctions into the products
of their electronic, rovibrational and angular parts, in
tegrating over the rotation angles and summing over M
transforms Eq. (1) into
1)D
J k
M
_{k} _{Ω} _{k} (φ, θ, 0)
(2)
×
J k
−M _{k}
1
P
M _{i} −Ω _{k}
J i
J _{k}
1
Q
where χ _{n} (R) are the rovibrational wave functions in
dexed by all the relativistic electronic channels that are
included in the model, d _{B}_{F} is the bodyﬁxed E1 tran
sition operator, J _{k} are the indexed continuum angu
J
lar momenta, D _{M}_{Ω} are Wigner rotation matrices, and
Q = Ω _{k} − Ω _{i} (a photodissociation transition is called
parallel if Q = 0 and perpendicular if Q = 1). The
polarization index P = 0 or P = ±1 if the photodisso
ciation light is polarized along or perpendicularly to the
quantization axis, respectively (such that P = M _{k} −M _{i} ).
B. Axial recoil approximation
At low photofragment energies, to obtain the correct
photodissociation cross sections it is crucial to calculate
the matrix elements χ
J k
(R)d _{B}_{F} χ _{n} _{i} _{J} _{i} _{Ω} _{i} (R) in Eq.
n k Ω k
(2). At high energies, these matrix elements become in
dependent of J _{k} [7] and can be simply factored out of Eq.
(2). This is equivalent to disregarding the photodissoci
ation dynamics. Under this axial recoil approximation,
the photodissociation cross section (2) reduces to
C. Twoparameter quantum mechanical model
Consider
the
photodissociation
of
0
+
u
(v, J _{i} , M _{i} )
molecules to the groundstate continuum with P = 0.
For odd
menta in
J _{i}
(which are the only allowed angular mo
0
+
u
due to quantum statistics), the allowed
angular momenta in the continuum are J = J _{i} − 1 and
J _{i} + 1, while M = M _{i} . Then the quantum mechanical
photofragmentation cross section can be expressed using
only two parameters, the channel amplitudesquared R
5
and the relative channel phase δ:
σ _{R}_{δ} (θ, φ) ∝
(4)
_{} ^{√} RY _{J} _{i} _{−}_{1}_{,}_{M} _{i} (θ, φ) + (−1) ^{Ω} ^{i} e ^{i}^{δ} ^{√} 1 − RY _{J} _{i} _{+}_{1}_{,}_{M} _{i} (θ, φ)
2
.
The (−1) ^{Ω} ^{i} factor correctly connects the sign of the δ
parameter to the phase shift between the continuum wave
functions, and Y _{J}_{M} are spherical harmonics which are
proportional to the angular wave functions for Ω = 0.
The expression for the R parameter can be derived by
comparing Eq. (4) to the traditional cross section (2).
We obtain
R =
_{} ^{} 2J _{i} − 1 ^{J} ^{i} ^{−} ^{1}
−M _{k}
^{1}
P
M i ^{} ^{} J _{i} − 1
J i
0
1
−Ω _{i}
i ^{χ}
J _{i}
Ω
J=J i−1
n k Ω k
(R)
d _{B}_{F}  χ _{n} _{i} _{J} _{i} _{Ω} _{i} (R)
2
×
J _{k} ={J _{i} −1,J _{i} +1}
_{} ^{} 2J _{k} + 1
J k
−M _{k}
1
P
M _{i} −Ω _{k}
J i
J _{k}
1
Q
i ^{χ}
J _{i}
Ω
_{k} _{Ω} _{k} (R) d _{B}_{F}  χ _{n} _{i} _{J} _{i} _{Ω} _{i} (R)
J k
n
2 ^{}
−1
.
(5)
The δ parameter is calculated as the phase shift diﬀerence
for the continuum wave functions, δ = δ _{J} _{i} _{+}_{1} − δ _{J} _{i} _{−}_{1} .
In the axial recoil limit Eq. (5) is simpliﬁed by setting
the matrix elements to a constant and canceling them, as
explained in Sec. III B. This results in values of R that
approach 0.5 with increasing J _{i} , while the phase shifts
become identical: δ _{J} _{i} _{+}_{1} → δ _{J} _{i} _{−}_{1} , yielding δ → 0 and
cos δ → 1.
D. Quasiclassical model
The
commonly
used
quasiclassical model for
photofragment angular distributions [5, 6] describes
the photodissociation cross section as
σ _{Q}_{C} (θ, φ) ∝ P _{i} (θ, φ) × [1 + β _{2} P _{2} (cos θ)] ,
(6)
J
where P _{i} (θ, φ) = D _{M} _{i} _{Ω} _{i} (φ, θ, 0) ^{2} is the probability den
i
sity of the initial molecular orientation. The anisotropy
parameter is β _{2} = 2 for parallel photodissociation tran
sitions (Q = 0) and β _{2} = −1 for perpendicular transi
tions (Q = 1). If the molecule is in a superposition
of J _{i} M _{i} Ω _{i} states, the initial probability density can be
generalized as [8]
P _{i} (θ, φ) =
^{}
J _{i} J M _{i} M
i
i
Ω i
J
D _{M} _{i} _{Ω} _{i} (φ, θ, 0)D
i
i
M
i
_{Ω} _{i} (φ, θ, 0).
J
(7)
Note that in Eq. (7) there are no cross terms for Ω _{i} .
The intuition behind Eq. (6) is that the photofragment
angular distribution follows the initial angular density of
the molecule, modiﬁed by the angular probability density
of the photon absorption.
E. Correspondence of the quasiclassical model to the axial recoil approximation
It was shown in Ref. [7] that for photodissociation light
polarized along the quantization axis, Eq. (3) reduces to
J
σ _{A}_{R} (θ, φ) ∝ D _{M} _{i} _{Ω} _{i}  ^{2} × [1 + β _{2} P _{2} (cos θ)] for initial states
i
with a single Ω _{i} [8]. We have conﬁrmed that this re
sult also holds for other polarizations of the photodisso
ciating light (if the light polarization is perpendicular to
the quantization axis, P _{2} (cos θ) should be replaced with
P _{2} (sin θ sin φ)).
Therefore the quasiclassical model (6) can be applied
in the axial recoil limit to obtain angular distribution
predictions that are identical to those of the quantum
mechanical model. However, an important assumption in
Refs. [7, 8] is that the photofragments are not identical
quantum particles. Deviations from this assumption are
discussed in Sec. V.
IV.
CROSSOVER FROM THE ULTRACOLD
REGIME TO THE AXIAL RECOIL LIMIT
We investigate the crossover from the ultracold quan
tum mechanical to the highenergy axial recoil regime
of photodissociation, as observed in Fig. 2. The intu
itive parametrization of Eq. (4) that accurately illus
trates twochannel photodissociation allows us to study
the crossover as it applies to only two parameters, R and
δ. These parameters can have a strong energy depen
dence in the ultracold regime while approaching their
axial recoil values at higher energies. In this Section,
we address the following questions: at what energy scale
does the outcome of photodissociation approach the ax
ial recoil limit; how does this energy scale depend on the
molecular binding energy (or vibrational level) and ro
tational angular momentum; and how well do the WKB
and semiclassical approximations model this crossover?
6
(a) 


v= 2
v= 3
v= 5
v= 0
v= 4




R 











ε/h (MHz) 

(b) 
V (r) 
J = 0
J = 2

trated for the 0
+
u
(−2, 1, 0) wave function in Fig. 1. The
continuum wave functions, as also shown in Fig. 1, are
similar at small internuclear separations but undergo a
relative phase shift at larger distances that correspond to
the bond length in initially very weakly bound molecules.
This can prevent photodissociation from reaching quasi
classical behavior at the expected energies. The energy
dependences for v = −4, −5 (bound by ∼ 10 ^{3} MHz) are
more regular, and we ﬁnd that all the deeper vibrational
levels starting from v = −10 (bound by ∼ 10 ^{5} MHz)
exhibit nearly identical behavior illustrated in Fig. 3(a)
by the deepest v = 0 level (bound by ∼ 10 ^{8} MHz). We
conclude that with the exception of asymptotic molecules
that are bound just below threshold, the R parameter is
independent of the vibrational quantum number. Since
the δ parameter is vindependent as discussed in Sec.
IV B, the expected anisotropy patterns are nearly inde
pendent of the binding energy, or v. We have checked
that for higher initial angular momenta J _{i} , the R param
eter is also independent of v except for the most weakly
bound molecules.
It is clear from Fig. 3(a) that R → 1 as ε → 0, indi
cating that the lowest partial wave (J = 0 in this case)
dominates the photofragment distribution at ultralow en
FIG. 3. (a) Quantum mechanical calculation of the amplitudesquared R parameter for twochannel photodisso
ciation as in Eq. (4). The initial molecular state is 0
+
u
(v, 1, 0)
with v = {−2, −3, −4, −5}, and P =
0. For comparison, the
most deeply bound v = 0 state is also shown. The axial recoil limit is indicated with a dashed line. (b) The allowed con tinuum wave functions with J = {0, 2} are plotted for very low energy ε/h = 0.5 MHz (ε/k _{B} = 24 µK). The negligible contribution of the J = 2 partial wave due to its repulsive ro tational barrier explains why R = 0 as ε → 0 in (a) and thus only the lowest allowed partial wave is populated in near threshold photodissociation.
ergies. This is caused by the suppression of lowenergy
scattering wave functions with higher angular momenta
at relevant internuclear separations, due to their rota
tional barriers. Figure 3(b) shows the J = 0 and J = 2
wave functions at the low continuum energy of ε/h = 0.5
MHz (ε/k _{B} = 24 µK), where J = 2 is strongly suppressed
at the relevant molecular bond lengths of 100 bohr.
The crossover from the ultracold, quantum mechanical
regime to the axial recoil limit is explored in Fig. 4 as
a function of the initial angular momentum J _{i} . The R
parameter is plotted as a function of energy for diﬀerent
odd J _{i} of initial states within the 0
+
u
and 1 _{u} electronic
manifolds, assuming M _{i} = P = 0. The calculations were
A. Approaching the axial recoil limit: Channel amplitude
The R parameter from Eq. (4) depends on both the
bound and continuum wave functions, as can be explicitly
seen from Eq. (5). Therefore, its behavior as a function
of continuum energy can depend on the vibrational state
of the molecule, potentially inﬂuencing the energy scale
at which the axial recoil limit is reached based on the
molecular binding energy. The plots in Fig. 3(a) show
the R parameter for several weakly bound initial states
+
0 u
(v, J _{i} = 1, M _{i} = 0) with v = {−2, −3, −4, −5}
as well
as for the most deeply bound state v = 0. The dashed
performed for the most deeply bound vibrational level
v = 0 to avoid any dependence on v. For all initial states,
R = 1 at threshold and decays to the axial recoil limit
(shown with horizontal lines) as a function of energy. In
each frame, a vertical line marks the rotational barrier
height corresponding to the largest partial wave in the
continuum, J _{i} + 1. As expected, the barrier height sets
the energy scale for the onset of the axial recoil regime.
This trend is evident despite the irregularities connected
to shape resonances for J = {4, 8, 20} (the J = 4 shape
resonance has been experimentally detected [9]).
line indicates the axial recoil limit.
For some of the shallow vibrational levels such as
B. Approaching the axial recoil limit: Channel phase and various approximations
v = −2, −3 (bound by ∼ 10 ^{2} MHz) the R parameter
shows strong energy dependence that does not closely
resemble that of more deeply bound states, as is also ob
served in Fig. 2. This is caused by the large spatial extent
( 100 bohr) of weakly bound wave functions, as illus
The δ parameter in Eq. (4) is the phase diﬀerence
between the continuum channel wave functions and is
independent of the initial state of the photodissociated
molecules. It is thus readily amenable to a range of ap
7
function is represented as an exponential with smoothly
varying amplitude and phase, and is a particularly use
ful method of ﬁnding δ without using the full quantum
mechanical treatment.
The continuum wave function corresponding to the
partial wave J has the asymptotic behavior
2 ,
Jπ
(8)
ψ _{J} (R) ∝ sin kR + δ _{J} −
where k ≡ ^{√} 2µε/ and µ is the reduced mass. An ana
lytical expression for δ _{J} is based on the WKB approxima
tion where the wave function is assumed to have the form
ψ _{J} (R) = e ^{i}^{S}^{(}^{R}^{)}^{/} ^{} . If only the lowestorder term in is
_{0} p(R)dR/ + π/4)
point and
p(R)
≡
^{} 2µ(E − V (R)) − J(J + 1)/R ^{2} in terms of the molec
ular potential V (R). The WKB approximation is most
applicable at higher energies and breaks down near the
turning point where p(R) ≈ 0. Based on the WKB ex
pression for ψ(R) and Eq. (8), the asymptotic phase shift
is
_{δ} WKB
J
=
R→∞
lim
R
1
R 0
p(R)dR +
π
4 − kR +
2
Jπ
.
(9)
kept in S(R), then ψ(R) ∝ sin( ^{}
where R _{0}
R
R
is the classical turning
A related approach to investigating the quantum
quasiclassical crossover is to consider the classical rota
tion of the molecule during photodissociation. We refer
to this method as the semiclassical approximation. Its
value is in the intuitive interpretation of the molecular
rotation angle during the bond breaking process and a
slightly faster computational convergence than the WKB
method. The angle of rotation γ between the initial
FIG. 4. Dependence of the amplitudesquared R parameter
from Eq. (4) on energy and on initial angular momentum
J _{i} . The initial molecular states are 0
+
u
(0, J _{i} , 0) (left column)
and 1 _{u} (0, J _{i} , 0) (right column), and P = 0. In each frame, the axial recoil limit is indicated by a horizontal line, and the highest rotational barrier in the continuum is shown by a dashed vertical line. For larger J _{i} , the axial recoil limit is approached at higher energies, but in all cases the crossover from the ultracold to the quasiclassical regime occurs at the energy scale set by the barrier height. The initial states with J _{i} = 3, 7, 19 exhibit shape resonance behavior.
molecular axis and the axis created by the scattered frag
ments is γ _{J} = ^{}
^{} J(J + 1)/[p(R)R ^{2} ]dR. This angle
_{0}
∞
R
is connected to the anisotropy parameter β _{2} by the rela
tion β _{2} = β _{A}_{R} P _{2} (cos γ), where β _{A}_{R} takes on the limit
ing values from the quasiclassical model [4–6] such that
β _{A}_{R} = {2, −1} for parallel and perpendicular photodis
sociation transitions, respectively. This semiclassical ap
proximation is more general than the quasiclassical model
which assumes that γ = 0.
The WKB method links the classical parameter γ to
the quantum mechanical phase δ _{J} through γ = −dδ _{J} /dJ
[20]. Further approximating the derivative by a diﬀerence
quotient yields the semiclassical approximation for the
phase shifts,
proximation techniques. We consider the WKB approxi
mation [19, 20] and the semiclassical approximation [21],
and investigate their applicability to photodissociation
across a wide span of energies as well as their conver
gence to the axial recoil limit.
The WKB approximation is a method of solving linear
diﬀerential equations with spatially varying coeﬃcients.
It is applicable to approximating the solutions to the
timeindependent Schr¨odinger equation, where the wave
δ _{J} − δ _{J} = (J ^{} − J) ^{γ} ^{J} ^{+} ^{γ} ^{J} ^{} .
2
(10)
The semiclassical approximation has been successfully
applied to diatomic molecule photodissociation at the en
ergy of several wave numbers above threshold [21], where
a deviation from the axial recoil limit was observed, yield
ing an agreement with quantum mechanical calculations
[22]. To summarize, the hierarchy of approximations,
8
FIG. 5. A comparison of the quantum mechanical, WKB, and
semiclassical methods for evaluating cos δ for a range of ini
tial molecular angular momenta J _{i} within the 0
+
u
electronic
manifold (here M _{i} = P = 0).
The phase angle δ quanti
ﬁes interference between the two allowed dissociation channels and plays a key role in the observed photofragment angular distributions. Both approximations approach the quantum mechanical result at increasingly higher energies that are set by the rotational barrier heights (dashed vertical lines). All methods recover the axial recoil limit (horizontal lines) at high energies, but this convergence is slower for the δ parameter than for the R parameter in Fig. 4.
from the most to the least exact, is (i) quantum me
chanical treatment, (ii) the WKB approximation, (iii) the
semiclassical model, and (iv) the axial recoil approxima
tion (which is generally equivalent to the quasiclassical
model with exceptions discussed in Sec. V and Ref. [8]).
Figure 5 shows a comparison of the quantum mechan
ical (Eq. (2)), WKB (Eq. (9)), and semiclassical (Eq.
(10)) approaches to calculating the δ parameter. This
phase angle is determined for the initial molecular states
J _{i} = {1, 3, 5, 7} within the 0
+
u
electronic manifold, as
suming photodissociation with M _{i} = P = 0. We ﬁnd
the applicability ranges of both approximation methods
to be similar, with the exception that the WKB approx
imation is more accurate for very low J _{i} . These ranges
are set by the barrier heights in the continuum, indicated
with dashed vertical lines. We also ﬁnd that all methods
converge to the axial recoil limit cos δ = 1. However, this
convergence happens more slowly than for the R param
eter in Fig. 4, and dominates the discrepancy between
the measured angular distributions and those expected
for axial recoil at energies beyond the barrier height, as
in Fig. 2 where the barrier height is only ∼ 10 MHz
but nonclassical behavior persists to ∼ 10 ^{3} MHz [23].
Furthermore, only the quantum mechanical approach is
sensitive to shape resonances, for example one that is
visible in Fig. 5 at ε/h ≈ 250 MHz for J _{i} = 7.
V. THE ROLE OF QUANTUM STATISTICS IN
PHOTODISSOCIATION
The equivalence of the quasiclassical approximation to
the axial recoil limit of the quantum mechanical model
holds for molecules in welldeﬁned Ω _{i} states under the
assumption that the photofragments are not identical
bosons or fermions. Section III E assumes that a molecule
with angular momentum J _{i}
= 0 can dissociate into the
{J _{i} −1, J _{i} , J _{i} +1} partial waves as allowed by E1 selection
rules.
Spin statistics can impose additional limitations on the
available continuum channels, as is the case for Sr _{2} and
many other molecules. In the electronic ground state, J
is always even for Sr _{2} composed of identical bosonic Sr
atoms and odd for Sr _{2} composed of identical fermions.
In this work, we use bosonic ^{8}^{8} Sr and therefore J is
even. This restriction only aﬀects Ω _{k} = 0 electronic
states. Furthermore, in cases where Ω _{i} = Ω _{k} = 0 or
M _{i} = M _{k} = P = 0, the properties of the 3j symbols in
Eqs. (2,3) naturally force the continuum J to be only
even or odd, making the axial recoil approximation in
sensitive to quantum statistics [24]. Here we extend Eq.
(3) to cases where quantum statistics must be considered,
and compare the results to the quasiclassical model (6).
In our experiment, this aﬀects photodissociation from the
1 _{u} states to the groundstate continuum, except in cases
where M _{i} = P = 0.
First we assume that the photodissociation light po
larization points along the quantization axis. The two
relevant cases are J = J _{i} (applicable, for example, to
fermionicSr dimers dissociating from odd J _{i} ) and J =
{J _{i} − 1, J _{i} + 1} (applicable to bosonicSr dimers pho
todissociating from odd J _{i} as in our experiments). As
shown in the Appendix, we adapt the axialrecoil cross
section σ _{A}_{R} (θ, φ) to both quantumstatistics restricted
cases, and ﬁnd the modiﬁed photodissociation cross sec
tion, σ _{Q}_{S}_{,} _{} (θ, φ). The result is analogous to the quasiclas
sical cross section σ _{Q}_{C} (θ, φ) in Eq. (6) but with the initial
J
angular probability density P _{i} (θ, φ) = D _{M} _{i} _{Ω} _{i} (φ,
i
replaced by
θ, 0) ^{2}
P _{i}_{,} _{} (θ, φ) = ^{} D _{M} _{i} _{,}_{1} (φ, θ, 0) − σ _{i} D _{M} _{i} _{,}_{−}_{1} (φ, θ, 0) ^{} ^{2} , (11)
J i
J i
where σ _{i} ≡ p _{i} (−1) ^{J} ^{i} is the spectroscopic parity of the
initial state and p _{i} is the parity with respect to space
ﬁxed inversion. For example, for 1 _{u} initial states, p _{i} = −1
for bosonicSr dimers and p _{i} = 1 for fermionicSr dimers.
Next we treat the case where the photodissociation
light polarization is perpendicular to the quantization
axis. Again spin statistics enforces either J = J _{i} or
J = {J _{i} − 1, J _{i} + 1}. As detailed in the Appendix, we
adapt σ _{A}_{R} (θ, φ) to both cases and ﬁnd the modiﬁed pho
todissociation cross section
σ _{Q}_{S}_{,}_{⊥} (θ, φ)
∝
(12)
J
cos ^{2} φ ^{} ^{} D _{M} _{i} _{,}_{1} (φ,
i
θ,
0)
+
σ _{i} D
J
_{M} _{i} _{,}_{−}_{1} (φ,
i
θ, 0) ^{}
2
+
J
cos ^{2} θ sin ^{2} φ ^{} D _{M} _{i} _{,}_{1} (φ,
i
θ,
J
0) − σ _{i} D _{M} _{i} _{,}_{−}_{1} (φ,
i
θ, 0) ^{}
2
.
9
FIG. 6. Quasiclassical and quantum mechanical calculations of the photofragment angular distributions do not generally converge in the axial recoil limit if the photofragments are identical particles. Experimental measurements always agree with the quantum theory [17]. Here the highenergy angular distributions are calculated with the quasiclassical (6) and quantum mechanical (3) models for the 1 _{u} initial states with (a) J _{i} = 1, M _{i} = {0, 1}, P = 1; (b) J _{i} = 3, M _{i} = {1, 2}, P = 0; and (c) J _{i} = 3, M _{i} = {0, 1, 2, 3}, P = 1.
This result is not analogous to the quasiclassical cross
section σ _{Q}_{C} (θ, φ).
The photodissociation cross sections modiﬁed by spin
statistics are not correctly described by the quasiclassical
model (6), even in the axial recoil limit of large contin
uum energies. This eﬀect is observable in experiments
as a deviation of highenergy photodissociation images
from the quasiclassical model for certain initial molecu
lar states and light polarizations [17].
Figure 6 illustrates ^{8}^{8} Sr _{2} photodissociation pathways
where photofragment angular distributions in the axial
recoil limit (3) cannot be described by the quasiclassi
cal model (6). In our experiment and in the ﬁgure, the
initial molecular states and laser polarizations that lead
to a nonquasiclassical axial recoil limit are (a) 1 _{u} (J _{i} =
1, M _{i} = {0, 1}, P = 1); (b) 1 _{u} (J _{i} = 3, M _{i} = {1, 2}, P =
0); and (c) 1 _{u} (J _{i} = 3, M _{i} = {0, 1, 2, 3}, P = 1). The
case
considered in the top row of Fig. 6(a) results in a dis
tribution that does not evolve with energy, since only a
single partial wave J = 2 is allowed in the continuum by
selection rules and hence the matrix elements in Eq. (5)
cancel for all energies [25]. Our experimental measure
ments of angular distributions always agree with quan
tum mechanical calculations [17]. For photodissociation
of molecules in analogous quantum states but composed
of identical fermionic isotopes we expect to observe diﬀer
ent distributions than in Fig. 6, and still diﬀerent results,
that are well described by the quasiclassical model, are
expected for mixed dimers.
VI.
CONCLUSIONS
We report photodissociation with ultracold ^{8}^{8} Sr _{2}
molecules in isolated internal quantum states, including
diﬀerent binding energies and angular momenta, where
we image and calculate the photofragment angular dis
tributions as a function of the kinetic energy in the con
tinuum, spanning from the ultracold to the quasiclassical
regime. At the higher energies that exceed the relevant
potential barriers, the distributions converge to a small
set of angular patterns that correspond to the axial recoil
limit where the molecule has no time to rotate during the
bond breaking process. In contrast, at the lower energies
the distributions exhibit strong variation with energy and
a dependence on the initial molecular state.
We utilize precise quantum state selection to ensure a
twochannel photodissociation outcome that can be de
scribed by only two parameters: an amplitude that has
a weak dependence on the initial vibrational quantum
number and a phase that only depends on the contin
uum wave functions. The amplitude converges to the
axial recoil limit signiﬁcantly faster than the phase, with
a slower convergence for very weakly bound initial molec
ular states. The WKB and semiclassical approximations
are shown to agree with the quantum mechanical model
at energies that exceed the barrier heights, and to cor
rectly approach the axial recoil limit. Finally, we ﬁnd
that in case of identical photofragments, the eﬀects of
bosonic or fermionic statistics can persist into the high
energy regime, in which case the axial recoil limit dis
agrees with the ubiquitous quasiclassical model.
The ability to access the crossover from the ultra
cold to the quasiclassical regime of photodissociation
was enabled by quantum state control of the molecules,
making it possible to populate exclusively lowangular
momentum states and thus access low partial waves in
the continuum that are strongly sensitive to quantum
eﬀects. As more experiments begin to probe cold and ul
tracold chemistry, including molecular photodissociation,
it is important to recognize the extent of the parameter
space where the processes are truly quantum mechanical.
Our work brings this insight into a longdebated issue
of the applicability of quasiclassical models to molecular
photodissociation, and elucidates which parameters limit
the convergence toward the highenergy limit.
ACKNOWLEDGMENTS
We acknowledge the ONR Grant No. N00014171
2246 and the NSF Grant No. PHY1349725. R. M. and
I. M. also acknowledge the Polish National Science Center
10
Grant No. 2016/20/W/ST4/00314 and M. M. the NSF
IGERT Grant No. DGE1069240.
Appendix: Axial recoil limit for identical photofragments
1. Light polarization along the quantization axis
With the assumptions {Ω _{i} = 1, Ω _{k} = 0, M _{k} = M _{i} }, as
for Sr _{2} photodissociation from the 1 _{u} states to the ground
state continuum with light polarization along the quan
tization axis (P = 0), the axialrecoil photodissociation
cross section (3) becomes
11
The second step of Eq. (A.4) uses Eq.
(A.2) as well as
the summation (ClebschGordan series)
(−1) ^{M} ^{k} ^{+}^{Ω} ^{k} (2J _{k} + 1)D
J k
= ^{} D _{−}_{M} _{k} _{−}_{Ω} _{k} (φ, θ, 0)
J k
J _{k}
= D
J i
M i Ω i
1
D P Q ^{.}
2. Light polarization normal to the quantization axis
With the assumptions {Ω _{i} = 1, Ω _{k} = 0, M _{k} = M _{i} ±1},
ground state continuum with light polarization normal
to the quantization axis (P = ±1), the axialrecoil cross
section (3) becomes
σ(θ, φ) ∝ ^{} (−1) ^{M} ^{k} (2J _{k} + 1)D
J _{k} P
a. Partial waves restricted to J _{i}
After restricting the partial waves to J _{k} = J _{i} , summing
over P, and evaluating the 3j symbols, the cross section
in Eq. (A.6) becomes
(A.7) 

2 

2 J i − cos θ D _{M} _{i} _{,}_{1} (φ, θ, 0) ,
2 J i − cos θ D _{M} _{i} _{,}_{−}_{1} (φ, θ, 0) 
× ^{} ^{} D _{M} _{i} _{+}_{1}_{,}_{0} (φ, θ, 0) ^{} (J _{i} − M _{i} )(J _{i} + M _{i} + 1) − D _{M} _{i} _{−}_{1}_{,}_{0} (φ, θ, 0) ^{} (J _{i} + M _{i} )(J _{i} − M _{i} + 1) ^{} .
J i
J i
We use the recursion formulas [26]
to transform Eq. (A.7) into
12
σ(θ, φ) ∝ ^{} ^{} D _{M} _{i} _{,}_{1} (φ, θ, 0) × (cos φ + i cos θ sin φ) − D _{M} _{i} _{,}_{−}_{1} (φ, θ, 0) × (cos φ − i cos θ sin φ) ^{} ^{} ^{2}
J
i
2
.
(A.10)
0) − D _{M} _{i} _{,}_{−}_{1} (φ, θ, 0) ^{} ^{} ^{2} + cos ^{2} θ sin ^{2} φ ^{} ^{} D _{M} _{i} _{,}_{1} (φ, θ, 0) + D _{M} _{i} _{,}_{−}_{1} (φ, θ, 0) ^{}
J i
J i
taking the sum over all partial waves and subtracting the
forbidden contributions,
_{k} _{0} (φ, θ, 0)
−M _{k} P M _{i} J ^{k} 1 J ^{i}
J k
1
J i
0
1
2
.
(A.11)
−(−1) ^{M} ^{k} (2J _{i} + 1)D _{M} _{k} _{0} (φ, θ, 0)
−M _{k} P M _{i} J ^{i} 1 J ^{i}
J _{i}
1
J _{i}
0
−1
second term on the righthand side of Eq. (A.11), respec
tively, we obtain the cross section
σ(θ, φ) ∝ ^{} D _{M} _{i} _{,}_{1} (φ, θ, 0) ^{} D _{1}_{,}_{−}_{1} (φ, θ, 0) + D _{−}_{1}_{,}_{−}_{1} (φ, θ, 0) ^{} +
2
2
=
1
J
D _{M} _{i} _{,}_{1} (φ, θ, 0) × (cos φ + i cos θ sin φ)
i
2 1 D _{M} _{i} _{,}_{−}_{1} (φ, θ, 0) × (cos φ − i cos θ sin φ) ^{}
= ^{} ^{} D _{M} _{i} _{,}_{1} (φ, θ, 0) × (cos φ + i cos θ sin φ) + D _{M} _{i} _{,}_{−}_{1} (φ, θ, 0) × (cos φ − i cos θ sin φ) ^{} ^{} ^{2}
J i
J i
(A.12)
J i
∝ cos ^{2} φ ^{} ^{} D _{M} _{i} _{,}_{1} (φ, θ, 0) +
D _{M} _{i} _{,}_{−}_{1} (φ, θ, 0) ^{} ^{} ^{2} + cos ^{2} θ sin ^{2} φ ^{} ^{} D _{M} _{i} _{,}_{1} (φ, θ, 0) − D _{M} _{i} _{,}_{−}_{1} (φ, θ, 0) ^{}
J i
J i
J i
2
.
[1] N. Balakrishnan, “Perspective: Ultracold molecules and the dawn of cold controlled chemistry,” J. Chem. Phys. 145, 150901 (2016). [2] H. Sato, “Photodissociation of simple molecules in the gas phase,” Chem. Rev. 101, 2687–2725 (2001). [3] R. N. Zare and D. R. Herschbach, “Doppler line shape of atomic ﬂuorescence excited by molecular photodissocia tion,” Proc. IEEE 51, 173–182 (1963). [4] R. N. Zare, “Photoejection dynamics,” Mol. Photochem. 4, 1–37 (1972). [5] S. E. Choi and R. B. Bernstein, “Theory of oriented symmetrictop molecule beams: Precession, degree of ori entation, and photofragmentation of rotationally state selected molecules,” J. Chem. Phys. 85, 150–161 (1986). [6] R. N. Zare, “Photofragment angular distributions from oriented symmetrictop precursor molecules,” Chem. Phys. Lett. 156, 1–6 (1989).
[7] T. Seideman, “The analysis of magneticstateselected angular distributions: a quantum mechanical form and an asymptotic approximation,” Chem. Phys. Lett. 253, 279–285 (1996). [8] J. A. Beswick and R. N. Zare, “On the quantum and quasiclassical angular distributions of photofragments,” J. Chem. Phys. 129, 164315 (2008). [9] M. McDonald, B. H. McGuyer, F. Apfelbeck, C.H. Lee, I. Majewska, R. Moszynski, and T. Zelevin sky, “Photodissociation of ultracold diatomic strontium molecules with quantum state control,” Nature 534, 122– 126 (2016). [10] M. McDonald, I. Majewska, C.H. Lee, S. S. Kondov, B. H. McGuyer, R. Moszynski, and T. Zelevinsky, “Con trol of ultracold photodissociation with magnetic ﬁelds,” Phys. Rev. Lett. 120, 033201 (2018). [11] W. Skomorowski, F. Pawlowski, C. P. Koch, and
13
R. Moszynski, “Rovibrational dynamics of the strontium
molecule in the A ^{1} Σ
+
u
c ^{3} Π _{u} ,
and a ^{3} Σ
+
u
manifold from
, stateoftheart ab initio calculations,” J. Chem. Phys.
136, 194306 (2012). [12] M. Borkowski, P. Morzy´nski, R. Ciurylo, P. S. Julienne,
M. Yan, B. J. DeSalvo, and T. C. Killian, “Mass scal
ing and nonadiabatic eﬀects in photoassociation spec troscopy of ultracold strontium atoms,” Phys. Rev. A 90, 032713 (2014). [13] B. H. McGuyer, C. B. Osborn, M. McDonald, G. Rein audi, W. Skomorowski, R. Moszynski, and T. Zelevinsky, “Nonadiabatic eﬀects in ultracold molecules via anoma
lous linear and quadratic Zeeman shifts,” Phys. Rev. Lett. 111, 243003 (2013). [14] B. H. McGuyer, M. McDonald, G. Z. Iwata, W. Sko morowski, R. Moszynski, and T. Zelevinsky, “Control of optical transitions with magnetic ﬁelds in weakly bound molecules,” Phys. Rev. Lett. 115, 053001 (2015). [15] B. H. McGuyer, M. McDonald, G. Z. Iwata, M. G. Tarallo, A. T. Grier, F. Apfelbeck, and T. Zelevinsky, “Highprecision spectroscopy of ultracold molecules in an optical lattice,” New J. Phys. 17, 055004 (2015). [16] B. H. McGuyer, M. McDonald, G. Z. Iwata, M. G. Tarallo, W. Skomorowski, R. Moszynski, and T. Zelevin sky, “Precise study of asymptotic physics with subradiant ultracold molecules,” Nature Phys. 11, 32–36 (2015). [17] S. S. Kondov, C.H. Lee, M. McDonald, B. H. McGuyer,
I. Majewska, R. Moszynski, and T. Zelevinsky,
“Crossover from the ultracold to the quasiclassical regime in stateselected photodissociation,” arXiv:1805.08850
(2018).
[18] G. Reinaudi, C. B. Osborn, M. McDonald, S. Ko
tochigova, and T. Zelevinsky, “Optical production of stable ultracold ^{8}^{8} Sr _{2} molecules,” Phys. Rev. Lett. 109, 115303 (2012). [19] M. S. Child, Molecular Collision Theory (Dover Publica tions, Mineola, New York, 2010). [20] E. E. Nikitin and S. Y. Umanskii, Theory of Slow Atomic Collisions (SpringerVerlag, Berlin Heidelberg, 1984). [21] E. Wrede, E. R. Wouters, M. Beckert, R. N. Dixon, and M. N. R. Ashfold, “Quasiclassical and quantum mechani cal modeling of the breakdown of the axial recoil approx imation observed in the near threshold photolysis of IBr and Br _{2} ,” J. Chem. Phys. 116, 6064 (2002). [22] In Ref. [21] the semiclassical model is referred to as “qua siclassical”. However, it is not equivalent to the quasiclas sical description from Refs. [5, 6], and is a more accurate approximation. [23] For the WKB images in Fig. 2, the axialrecoil value of R was used, since R approaches the highenergy limit faster than δ. [24] For M _{i} = P = 0, quantum statistics (when combined with the properties of the 3j symbol in Eqs. (2,3)) can cause a normally allowed photodissociation path way to become forbidden, for example for initial states 1 _{u} (v, J _{i} , 0) with even J _{i} in bosonicSr dimers. Such tran sitions, however, can be enabled with weak magnetic ﬁelds [9, 10]. [25] Two other conﬁgurations with this property, 1 _{u} (J _{i} = 1, M _{i} = 1, P = 0) and 1 _{u} (J _{i} = 3, M _{i} = 3, P = 0), are not shown. [26] D. A. Varshalovich, A. N. Moskalev, and V. K. Kher sonskii, Quantum Theory of Angular Momentum (World Scientiﬁc, Singapore, 1988).